|
id | OptGP () |
|
id | Xnew () |
|
hypVar, Xnew, Ynew, K_M, Opt.GP | NLikelihood () |
|
see Jones paper | lb (h1+h2) |
|
| ub (h1+h2) |
|
Xnew | size () |
|
| warning ('Covariance matrix in Nlikelihood is not positive semi-definite') end % invK = 2*sum(log(abs(diag(CH)))) |
|
dNLL_f | cov (i) |
|
dNLL_f | lik (i) |
|
id | nSpectralpoints () |
|
Posterior | sample (function) according to theta f |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY OR CONSEQUENTIAL | DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF % SUBSTITUTE GOODS OR SERVICES;LOSS OF USE, DATA, OR PROFITS;OR BUSINESS % INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY OR CONSEQUENTIAL WHETHER IN STRICT OR | TORT (INCLUDING NEGLIGENCE OR OTHERWISE) % ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE |
|
| error (nargoutchk(0, 1, nargout)) |
|
|
file Train_GP_and_return_hyperparameters m brief Trains a Gaussian processe and returns | hyperparameters |
|
file Train_GP_and_return_hyperparameters m brief Trains a Gaussian processe and returns RWTH Aachen University | n |
|
file Train_GP_and_return_hyperparameters m brief Trains a Gaussian processe and returns RWTH Aachen University Xiaopeng | Lin |
|
file Train_GP_and_return_hyperparameters m brief Trains a Gaussian processe and returns RWTH Aachen University Xiaopeng Daniel | Grothe |
|
file Train_GP_and_return_hyperparameters m brief Trains a Gaussian processe and returns RWTH Aachen University Xiaopeng Daniel and Alexander Mitsos date January | function [Opt] |
|
Scale | Variables [xScaled, yScaled] = ScaleVariables(X,Y,lb,ub) |
|
Training of GP for | j |
|
end Calculate covariance matrix and inverse of covariance matrix for | i |
|
end return Artur M Schweidtmann and Alexei | Lapkin |
|
Set initial hyperparameters | h1 = Opt.GP.h1 |
|
number of hyperparameters from covariance | h2 = Opt.GP.h2 |
|
| K_M = zeros(n,n*D) |
|
Define bounds | lb = ones(h1+h2,1) * log(sqrt(10^(-3))) |
|
see Jones paper | ub = ones(h1+h2,1) * log(sqrt(10^(3))) |
|
| bounds = [lb,ub] |
|
opts | maxevals = Opt.GP.fun_eval*(h1+h2) |
|
opts | maxits = 100000*(h1+h2) |
|
opts | maxdeep = 100000*(h1+h2) |
|
opts | showits = 0 |
|
Defintion of options for global | search [~, x0] = Direct(obj_fun,bounds,opts) |
|
Defintion of options for fmincon solver LSoptions | Algorithm = 'interior-point' |
|
LSoptions | DerivativeCheck = 'off' |
|
LSoptions | TolCon = 1e-12 |
|
LSoptions | Display = 'off' |
|
LSoptions | Hessian = 'bfgs' |
|
LSoptions | TolFun = 1e-12 |
|
LSoptions | PlotFcns = [] |
|
LSoptions | GradConstr = 'off' |
|
LSoptions | GradObj = 'on' |
|
LSoptions | TolX = 1e-14 |
|
LSoptions | UseParallel = 0 |
|
Solve optimization problem | hypResult = fmincon(obj_fun.f,x0,[],[],[],[],lb,ub,[],LSoptions) |
|
Return optimal hyperparameters Opt GP hyp | cov = hypResult(1:h1) |
|
Opt GP hyp | lik = hypResult(h1+1:h1+h2) |
|
| OptGPhyp = Opt.GP.hyp |
|
Opt | GP = OptGP |
|
type of Martern else | d = 1 |
|
number of hyperparameters from likelihood | hyp = Opt.GP.hyp |
|
| ell = exp(hyp.cov(1:D)) |
|
| sf2 = exp(2*hyp.cov(D+1)) |
|
| K = zeros(n,n) |
|
| expnK = exp(-sqrtK) |
|
| sqrtK = [] |
|
end if Opt GP | t = sqrtK |
|
| m = (1 + t).*expnK |
|
This guarantees a symmetric matrix Calculate inverse of covariance matrix try | CH = chol(K) |
|
| invK = CH\(CH'\eye(n)) |
|
Calculate hyperperpriors | logprior = 0 |
|
| dlogpriorcov = zeros(1,h1) |
|
end | dlogpriorlik = zeros(1,h2) |
|
Gradient calculation if | nargout |
|
end | c = invK*Ynew |
|
| b = invK* dK |
|
end | dNLL = [dNLL_f.cov' |
|
end return function | f |
|
| sn2 = exp(2*Opt.hyp.lik) |
|
else | W = randn(nSpectralpoints,D) .* repmat(1./ell', nSpectralpoints, 1) |
|
Calculation of phi | phi = sqrt(2 * sf2 / nSpectralpoints) * cos(W * Xnew' + repmat(b, 1, n)) |
|
Sampling of theta according to phi | A = phi * phi' + sn2 * eye(nSpectralpoints) |
|
| invA = invChol(A) |
|
| mu_theta = invA*phi*Ynew |
|
| cov_theta = sn2*invA |
|
| theta = mvnrnd(mu_theta,cov_theta)' |
|
return function | v |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary | forms |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without | modification |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are | met |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright | notice |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED | WARRANTIES |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED | INCLUDING |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED | TO |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY | DIRECT |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY | INDIRECT |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY | INCIDENTAL |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY | SPECIAL |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY | EXEMPLARY |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY OR CONSEQUENTIAL WHETHER IN | CONTRACT |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY OR CONSEQUENTIAL WHETHER IN STRICT | LIABILITY |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY OR CONSEQUENTIAL WHETHER IN STRICT OR EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE HYPERVOUME Hypervolume indicator as a measure of Pareto front estimate | V |
|
return function Yi Cao All rights reserved Redistribution and use in source and binary with or without are permitted provided that the following conditions are this list of conditions and the following disclaimer *Redistributions in binary form must reproduce the above copyright this list of conditions and the following disclaimer in the documentation and or other materials provided with the distribution THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS AS IS AND ANY EXPRESS OR IMPLIED BUT NOT LIMITED THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY OR CONSEQUENTIAL WHETHER IN STRICT OR EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE HYPERVOUME Hypervolume indicator as a measure of Pareto front estimate paretoGroup Version by Yi Cao at Cranfield University on April | Example |
|
upper bound of the data set | r =max(F) |
|
Approximation of Pareto set | P =paretofront(F) |
|
https | __pad0__ |
|
https | nargin |
|